Shape analysis and classification of masses in mammographic images using neural networks

Shape analysis of masses in mammographic images includes representation of mass contour and shape factors which are important features for distinguishing between benign and malignant masses. Three shape factors, namely compactness, moments and Fourier descriptors were calculated and used for the classification. Classification was performed using two types of neural networks: single layer and multilayer perceptron. Area under the ROC curve of A
_z
= 0.9528 was achieved using perceptron with 1000 epochs and using all shape factors. Multilayer perceptron with 1000 epochs and all shape factors achieved better classification results and area under the ROC curve of A
_z
= 0.9988.